Fostering Learning with Facial Insights: Geometrical Approach to Real-Time Learner Engagement Detection

The global shift in the education paradigm demands technology-enabled learning platforms that monitor the performance and engagement of learners. In this paper, a new method of detecting learner engagement levels utilizing geometric angle features obtained from face images captured through a webcam attached to the learning device is proposed. Distinct ten angles resulting from lines joining seventeen landmark points serve as the feature set and employ machine learning algorithms for classification. The algorithms used for the classification purpose include Random forest, K-Nearest Neighbour, and Support Vector Machine. Random Forest outperforms the other classifiers in distinguishing between the two engagement levels being tested. It achieves the highest accuracy of 90% in distinguishing low and high engagement levels of students. The 3-Level classification model is also comprehensively compared with other state-of-the-art methods and found a notable accuracy than the deep learning models MobileNet and VGG16. This indicates that the proposed method can efficiently detect and output students’ engagement levels which provides real-time support for teachers in classrooms and helps students while undergoing self-paced E-learning.